This study is a prospective, multicenter cohort of persons with HTLV-I and -II infections, which were detected at the time of attempted blood donation at five U.S. blood centers and comparable HTLV–seronegative donors. Details of the cohort enrollment and follow-up procedures have been published previously (16,17). The study protocol was approved by the USCG committe on Human Research and by IRBs at other participating institituions

We determined HTLV serostatus by obtaining enzyme immunoassay test results followed by confirmatory Western blot. A central laboratory performed HTLV-I versus -II typing with a type-specific serologic assay, polymerase chain reaction (PCR), or both, as previously described (18). Unequivocal results from the type-specific serologic assay correlated well with those from the polymerase chain reaction assay. All participants were seronegative for HIV when baseline test were performed. For most participants, hepatitis C virus antibody status was not available at the time of enrollment.

Each visit with a study nurse consisted of an interviewer-administered health history questionnaire and phlebotomy of blood for complete blood count and other studies. Selected diagnoses (cancer, neurologic and autoimmune conditions) reported on the questionnaire triggered requests for confirmatory medical records. We included nine conditions or diseases (pneumonia, acute bronchitis, bladder or kidney infection, arthritis, hypertension, asthma, cancer, diabetes, and thyroid disease) and eight symptoms (trouble walking, climbing, or rising from chair; incontinence; pre- or post-void urgency; lymphadenopathy; night sweats; weight loss; foot paresthesias; and impotence [males]) in the data analysis.

We used the Kaplan-Meier product limit method to calculate the unadjusted probability of disease-free survival during the study period for each disease outcome by HTLV status. Survival time was defined as the number of days from the baseline visit until the date that an adverse health outcome was first diagnosed or the end of observation. We performed the log-rank test to assess the differences in disease‑free survival time (days) between HTLV-seronegative participants and HTLV‑I– or HTLV-II– infected participants, respectively.

To adjust for possible confounding factors, we performed multivariable analysis with HTLV status as an independent variable, survival time as a dependent variable, and possible confounding factors as covariates. In constructing the survival analysis models, we considered a number of covariates, which are described as follows: demographic variables (forced into all models), education, smoking history (pack-years, forced into the models for bronchitis and pneumonia), alcohol consumption, blood center, community versus autologous donation type, injection drug use (except in models for arthritis, hypertension, cancer, neurologic and urologic symptoms), parity (in models for urinary symptoms, bladder and kidney infections, and in females only), and number of sex partners (in model for bladder and kidney infections only). Using a backward selection process, all these covariates were added to the initial model, but only covariates with significant independent associations themselves (p < 0.05), or which substantively modified HTLV incidence ratio, were retained in the final model. We did not include interaction terms because similar analyses of data from prior cohort visits had indicated the absence of significant interaction.

To examine the differences in the cumulative number of episodes of pneumonia, bronchitis, and bladder or kidney infection by HTLV status, we used the negative binomial model, a generalization of the Poisson model, to compute incidence rate ratios (RR) with 95% confidence interval (CI) for each of these outcomes. This model took into account that the recurrence of the disease in a participant may be associated with both the overall disease incidence and its previous occurrence in that participant. We also adjusted for possible confounding factors using the same modeling strategy as for the survival analyses.

For the analysis of symptoms, we calculated unadjusted and adjusted odds ratio (OR) with 95% CI for any occurrence of each symptom by HTLV status by using logistic regression models. To adjust for possible confounding, we added other possible confounding factors as covariates to the models by using the same approach as in the survival analyses. All analyses were performed using Statistical Analysis System (SAS) software (19).

During the initial visits in 1990 through 1992, we enrolled 154 HTLV-I, 387 HTLV-II, and 799 HTLV seronegative persons. Two HTLV-I participants were excluded from this analysis because they did not complete the screening physical examination at the initial visit. The baseline characteristics of the study population are given in Table 1. The HTLV groups and seronegative participants were comparable with respect to age, sex, race or ethnicity, blood center visited, and type of blood donation (autologous versus allogeneic), except for slightly higher proportions of blacks among the HTLV-I group and Hispanics among the HTLV-II group. The HTLV-II group had the lowest socioeconomic status, as indicated by educational attainment and income, and the HTLV-I group had intermediate status. Pack-years of cigarette smoking and amount of alcohol intake were higher in the HTLV-II group. Consistent with the recognized epidemiologic risk factors for HTLV-I and -II infection (3,4), the HTLV-I and -II groups had more lifetime sexual partners than seronegative participants, and almost 24% of the HTLV-II participants had a lifetime history of injection drug use, although only 4.4% of the HTLV-II group admitted to current injection drug use.

We present data through the third study visit in 1995 through 1996. Median follow-up time was 4.3 years for all 1,338 participants, including those with no follow-up. Median follow-up did not differ by HTLV status and was 4.4 years for the HTLV-I group, 4.3 years for the HTLV-II group, and 4.4 years for the HTLV-seronegative group.

Compared to seronegative persons, HTLV-I–infected persons were more likely to have a new diagnosis of bladder or kidney infection (p = 0.009) and arthritis (p = 0.0002) (Figure). Among the HTLV-I–infected persons, two had asthma; an insufficient number to test the difference relative to seronegative persons. The HTLV-I participants showed no statistically significant differences in the incidence of pneumonia, acute bronchitis, and hypertension, cancer, diabetes, and thyroid disease (data not shown).

The number of incident cases diagnosed (limited to one case per person) and the unadjusted and adjusted incidence ratios (IR) for several diagnoses are given in Table 2. Compared with results for seronegative persons, and after multivariable adjustment for relevant confounding variables, HTLV-I infection was associated with bladder or kidney infection (IR 1.82, 95% CI 1.19 to 2.77) and arthritis (IR 2.84, 95% CI 1.51 to 5.33). The risks of developing pneumonia, acute bronchitis, hypertension, and cancer were not significantly increased. Too few cases of asthma (n = 2), thyroid disease (n = 1), and diabetes mellitus (n = 5) existed among the HTLV-I participants to perform survival analysis.

To further investigate the occurrence of infectious diseases, we also analyzed the incidence density of three infectious diseases by HTLV status, whether or not each diagnosis was a first or recurrent case (Table 3). In the incidence density analysis (Table 3), an average of 1.75 cases of bladder or kidney infection occurred per HTLV-I participant over the 4.4-year median follow-up time, compared with 0.63 per seronegative participant over the same period. The unadjusted and adjusted RRs for the HTLV-I group were significantly greater than unity for bladder or kidney infection. HTLV-I participants had increased prevalence rates of neurologic symptoms, self-reported lymphadenopathy, and night sweats, but they reported weight loss no more frequently than did HTLV-seronegative persons (Table 4).

Disease-free survival curves for selected medical diagnoses, by HTLV status, are represented in the Figure. Compared to seronegative persons, HTLV-II–infected participants were more likely to experience acute bronchitis (p < 0.0001), bladder or kidney infection (p = 0.0008), arthritis (p = 0.0003), and asthma (p = 0.0007); the likelihood of acquiring pneumonia was increased but not significantly (p = 0.08). Differences between the HTLV-II and HTLV-seronegative participants in the incidence of cancer, hypertension, diabetes, or thyroid disease were not statistically significant.

The number of incident cases diagnosed (limited to one case per participant), and the unadjusted and adjusted IRs, for several diagnoses are given in Table 2. Compared with results for seronegative participants, and after adjusting for confounding variables, HTLV-II was associated with acute bronchitis (OR 1.68, 95% CI 1.24 to 2.29), bladder or kidney infection (OR 1.55, 95% CI 1.14 to 2.11), arthritis (OR 2.66, 95% CI 1.58 to 4.45), and asthma (OR 3.28, 95% CI 1.57 to 6.84). The association between HTLV-II infection and pneumonia was of borderline statistical significance (IR 1.82 , 95% CI 0.98 to 3.38). IRs for hypertension and cancer were not increased for HTLV-II participants. Too few cases of thyroid disease (n = 7) and diabetes mellitus (n = 7) were found among the HTLV-II participants to perform survival analysis.

In the incidence density analysis of recurrent infections, an average of 0.21 cases of pneumonia, 1.10 cases of acute bronchitis, and 1.25 cases of bladder or kidney infection occurred among HTLV-II participants during their 4.3-year median follow-up time (Table 3). The corresponding incidence densities for the seronegative group were 0.08, 0.59, and 0.63, respectively, over their 4.4 year median follow-up time. Both unadjusted and adjusted RRs, calculated by using negative binomial modeling, were significantly greater than unity for all three diseases (Table 3). Compared with seronegative participants, HTLV-II participants had increased prevalence rates of several neurologic, lymphatic, and constitutional symptoms, but the prevalence of impotence (males) was not significantly increased (Table 4).